Public sector has unprecedented number of computer platform choices

This industry has more than its share of myth-makers and FUD (fear, uncertainty and doubt) rakers. During the early 1980s with DEC, Honeywell and Data General as key minicomputer makers, there was at least a fairly clear cleavage, both in the private and public sectors: large business applications of any size went to the mainframe. Some departmental minis ran localized business applications, but most ran operational, dispatch, command-control and supervisory/monitoring types of applications. The biggest exaggerations then were that the new superminis, most notably the VAX-11-780 and some of the specialist simulation machines, such as Perkin-Elmers, were “approaching mainframe class.” They were not — and several major Goverment of Canada projects in Ottawa crashed because of it. In the later 1980s we saw the emergence of the practical manifestation of Unix, although in many flavours. Practical, fully relational Unix-based databases soon followed, most notably Informix and Oracle. What only a few of us realized at that time was that these new 32-bit RISC machines could indeed replace at least some midframes, could be used as local workgroup systems and could also be installed in operational and even in mobile/field/embedded implementations. Meanwhile, the FUD machine was telling us that running business applications over Unix and Oracle would never be practical, much less reliable — we should stick to MVS and DB/2. There were some problems. For example, the DND ran a project called Base ADP, which took midrange equipment to the base level, but that did not end happily. There were also successes. I led a wonderful group at Transport Canada who proved every one of the above assertions to be false. I also certified five Unix vendors with a single Unix API set and even trained/certified para-professionals to administer them in userland. Our five winning vendors (AT&T, BULL, DEC, HP and Unisys) were at the time — all things considered — the finest Unix midrange system vendors in the world and our installed cost of MIPS was approximately 1/15th that of a mainframe.
However, a new argument then arose that CPU and I/O capacity was growing so rapidly that soon everyone would soon have a mainframe on their desk and the midrange and mainframe tiers would disappear. This myth also had a kissing cousin, which said that all applications would soon be pure client-server with the mainframe on the back end and the PC on the front end. This line of argument had two fallacies. First, it ignored the issue of architectural placement — the overall arrangement of a group of systems, what the applications each machine do and how they relate to those on other platforms in the group. Second, there was the issue of how much power is in a mainframe. My Texas Instruments BA-III business calculator has more power than some fridge-sized computers of the 1970s, while most teenagers today carry notebooks or PDAs with more power than the earliest IBM mainframes. The truth was that as basic CPU power grew, all of PC, midrange, midframe and mainframe power ratings grew apace. So, by the early 1990s, Unix was established as a viable environment for business systems in government, and departments that needed operational, real-time or mobile/field/embedded capabilities, such as Transport, Defence, Environment, etc. could use the same CPUs in desktop workstations, workgroup systems and out in the field. The trend toward putting small systems in userland, which we and a few others across North America began, accelerated rapidly towards the mid-1990s as Microsoft fielded NT on increasingly powerful Intel CPUs, eventually multi-processor systems, which by mid-decade were making clear inroads at the bottom end of the Unix RISC market. While Unix drove everything except OS/400 off the midrange tier (by 1995 all other non-UNIX midrange computer product lines were declining or dead) it was, in turn, assaulted by NT moving up-tier. The choices for the public sector IT manager have never been greater, from the smallest micro-based server to the most powerful super-server, but we continue, it would seem, to be dogged by myth.